|
|
Received: 3 February 2017 Revised: 13 March 2017 Accepted: 27 March 2017 DOI: 10.1002/ece3.3006
ORIGINAL RESEARCH
The role of landscape characteristics for forage maturation and nutritional benefits of migration in red deer Atle Mysterud1
| Brit Karen Vike1 | Erling L. Meisingset2 | Inger Maren Rivrud1
1 Department of Biosciences, Centre for Ecological and Evolutionary Synthesis, University of Oslo, Blindern, Norway 2
Department of Forestry and Forestry Resources, Norwegian Institute of Bioeconomy Research, Tingvoll, Norway Correspondence Atle Mysterud, Centre for Ecological and Evolutionary Synthesis, Department of Biosciences, University of Oslo, Blindern, Norway. Email:
[email protected] Funding information Norges Forskningsråd, Grant/Award Number: 230275
Summary Large herbivores gain nutritional benefits from following the sequential flush of newly emergent, high-quality forage along environmental gradients in the landscape, termed green wave surfing. Which landscape characteristics underlie the environmental gradient causing the green wave and to what extent landscape characteristics alone explain individual variation in nutritional benefits remain unresolved questions. Here, we combine GPS data from 346 red deer (Cervus elaphus) from four partially migratory populations in Norway with the satellite-derived normalized difference vegetation index (NDVI), an index of plant phenology. We quantify whether migratory deer had access to higher quality forage than resident deer, how landscape characteristics within summer home ranges affected nutritional benefits, and whether differences in landscape characteristics could explain differences in nutritional gain between migratory and resident deer. We found that migratory red deer gained access to higher quality forage than resident deer but that this difference persisted even after controlling for landscape characteristics within the summer home ranges. There was a positive effect of elevation on access to high-quality forage, but only for migratory deer. We discuss how the landscape an ungulate inhabits may determine its responses to plant phenology and also highlight how individual behavior may influence nutritional gain beyond the effect of landscape. KEYWORDS
elevation, movement ecology, normalized difference vegetation index, partial migration, seasonality, ungulates
1 | INTRODUCTION
new growth, starting at low elevations (or latitudes), and moving to-
Migration between separate seasonal home ranges is a common phe-
the green wave (van der Graaf, Stahl, Klimkowska, Bakker, & Drent,
nomenon across animal taxa in many ecosystems all over the globe
2006; Merkle et al., 2016). Early phenological growth stages of plants
(Bauer & Hoye, 2014; Bolger, Newmark, Morrison, & Doak, 2008;
have higher nutritional quality due to high cell soluble content and
Fryxell, Greever, & Sinclair, 1988). At northern latitudes with strong
low levels of defense compounds (Van Soest, 1994). The basis for
seasonality, large migratory herbivores move from winter to summer
the forage maturation hypothesis is that large migratory herbivores
ranges when snow gradually melts in the spring and new vegetation
will preferentially follow phenological gradients or green waves in
of high nutritional quality emerges. Environmental gradients in the
order to maximize access to the optimal combination of quality and
landscape cause a predictable sequence of a green flush of fresh,
quantity of forage (Albon & Langvatn, 1992; Fryxell & Sinclair, 1988;
ward higher elevations (or latitudes), a phenomenon referred to as
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. © 2017 The Authors. Ecology and Evolution published by John Wiley & Sons Ltd. 4448 | www.ecolevol.org
Ecology and Evolution. 2017;7:4448–4455.
|
4449
MYSTERUD et al.
Hebblewhite, Merrill, & McDermid, 2008), ultimately resulting in increased body growth, reproductive rates, and survival rates (White, 1983). There are now several studies providing empirical support of the
2 | MATERIAL AND METHODS 2.1 | Study area
forage maturation hypothesis, demonstrating that herbivores utilize
The data were derived from four counties on the west coast of Norway:
spatial variation in the onset of plant growth to enhance the dura-
Hordaland, Sogn & Fjordane, Møre & Romsdal, and Sør-Trøndelag,
tion of access to newly emergent, high-quality plants (Bischof et al.,
the core area for red deer in Norway in terms of historical distribu-
2012; Hebblewhite et al., 2008; Merkle et al., 2016; Searle, Rice,
tion and population density. The study area has a diverse topogra-
Anderson, Bishop, & Hobbs, 2015). These studies have provided
phy, from flat coastal areas to steep fjord landscapes and mountains.
support for several predictions from the forage maturation hypoth-
The temperature and snow depth increase from the coast to inland
esis: (1) that migratory individuals gain access to a higher quality diet
(Mysterud, Yoccoz, Stenseth, & Langvatn, 2000). The vegetation is in
than resident individuals (Bischof et al., 2012; Hebblewhite et al.,
the boreonemoral zone for the most part, with a small proportion of
2008), (2) that migratory individuals gain access to newly emergent
Sør-Trøndelag in the southern boreal zone and a small proportion of
plants by migrating between separate ranges compared to remain-
Hordaland in the nemoral zone (Abrahamsen et al., 1977). The natural
ing in their winter ranges (Bischof et al., 2012), and (3) that herbi-
forests are characterized by Scots pine (Pinus sylvestris) and deciduous
vores actually follow the green wave (Merkle et al., 2016). However,
trees such as birch (Betula spp.) and gray alder (Alnus incana). Planted
there has been limited effort to relate the individual variation in
Norway spruce (Picea abies) has a patchy distribution across the study
access to newly emergent plants to the landscape characteristics
area. Agricultural areas are mainly located on flatter ground near
that cause environmental gradients in the onset and development
the coast or on valley floors. The cultivated land is mostly meadows
of plant growth, such as latitude, distance to coast, elevation, slope,
and pastures for grass production (Lande, Loe, Skjærli, Meisingset, &
and aspect. At high elevations and further inland, the snow cover
Mysterud, 2014). Some grains (Hordeum vulgare and Avena sativa) are
remains for a longer time in the spring, and together with lower tem-
produced in the warmest and most fertile areas, particularly in Sør-
peratures, this causes delayed forage development during the sum-
Trøndelag county.
mer. The summer ranges of red deer (Cervus elaphus) inland and at higher elevations have higher forage quality in late summer (Albon & Langvatn, 1992). The forage quality at migration stop-over sites
2.2 | Red deer data
was positively correlated with elevation for mule deer (Odocoileus
We used GPS data from 346 collared red deer that were followed
hemionus) (Sawyer & Kauffman, 2011). We may also expect delayed
along the west coast of Norway in the period from 2004 to 2014.
phenology at sites with a more northerly aspect and along flat ter-
Subsets of the data have been used previously (Bischof et al., 2012;
rain; such sites allow snow to accumulate, delaying plant growth in
Mysterud et al., 2011; Rivrud et al., 2016). The procedure used to
the spring. It remains largely an open question which landscape vari-
collar the red deer has been approved by the Norwegian Animal
ables other than elevation underlie the most beneficial phenological
Research Authority, and the chemical immobilization and marking
gradient for ungulates, yielding the highest access to high-quality
methods follow standard protocols (Sente et al., 2014). Adult deer
forage.
(females ≥ 1.5 years; males ≥ 2.5 years) were marked with GPS col-
The aim of this study was to test how landscape characteristics,
lars (Tellus from Followit, Sweden, and GPS ProLite from Vectronic,
habitat type, and individual home range characteristics predict the
Germany) and weighed to the nearest 0.5 kg. The collars were set to
access of 346 individual GPS-marked red deer to newly emergent
download a position every hour or every second hour. As some indi-
plants in four populations in the variable landscapes of Norway. The
viduals were followed for more than 1 year, we only used data from
combination of GPS-based telemetry and satellite images measuring
the first recorded season per individual to avoid pseudoreplication.
the greenness of the vegetation, such as the normalized difference
Locations recorded during the first 24 hr after marking were removed,
vegetation index (NDVI), now allow us to explore such relationships in
and the raw data were screened for outliers (Bjørneraas, Van Moorter,
detail (Bischof et al., 2012). We aim to test the following predictions
Rolandsen, & Herfindal, 2010).
from the forage maturation hypothesis: (P1) Migratory animals have
Space use tactics were determined using the Net-Square
access to higher quality forage (higher cumulative instantaneous rate
Displacement (NSD) technique (Bunnefeld et al., 2011), but modified
of growth, CIRG) than resident animals. (P2) The variation in landscape
so that individual fit was assessed manually, as in our previous work
characteristics in summer home ranges, such as elevation, distance to
(Bischof et al., 2012; Mysterud et al., 2011; Rivrud et al., 2016). We
fjord, aspect, and slope, causes variation among individuals in terms of
only included individuals classified as migrants (n = 190) or residents
their access to high-quality forage (CIRG). (P3) Variation in landscape
(n = 156).
characteristics in the summer home ranges explains the differences between migratory and resident deer in terms of access to high-quality forage (CIRG). We further tested whether the effects of landscape
2.3 | Home range characteristics
characteristics affected resident and migratory deer in the same way
As spring migrations in Norwegian red deer are rapid and closer to jump-
(i.e., if there were interactions).
ing than surfing in the wave use continuum (Bischof et al., 2012), we
|
MYSTERUD et al.
4450
used the individual’s summer home range as a basis for the demarcation
Hansen, & Lawrence, 2016; Hamel, Garel, Festa-Bianchet, Gaillard,
of landscape characteristics. Ninety-five percent utilization distribution
& Côté, 2009). Data on the NDVI were extracted by downloading
home ranges were calculated using fixed-kernel density estimation in
images covering Norway derived from the satellite MODIS TERRA
the R package adehabitat (Calenge, 2006). The reference method was
(MOD13Q1) and available from the NASA Land Processes Distributed
used to calculate the smoothing factor, h, for each individual. For resi-
Active Archive Center website (https://lpdaac.usgs.gov/data_access/
dent red deer, GPS fixes from 1 April to 31 August were used to match
daac2disk). The spatial resolution of these images is 250 m, and the
with the growing season used in our analysis. Summer home ranges
temporal resolution is 16 days. For each 16-day period, the images
for migratory deer were calculated using GPS fixes from the time they
were merged and subsampled using the MODIS reprojection Tool
reached the summer ranges until they departed back to winter ranges.
v.4.1 (https://lpdaac.usgs.gov/tools/modis_reprojection_tool).
A range of landscape covariates was extracted from the individual
In accordance with our earlier study (Bischof et al., 2012), we ex-
home ranges by overlaying the home range polygons on the landscape
tracted information about the instantaneous rate of green-up (IRG),
maps, and the means of all pixel values within the home ranges were
measuring the speed of the plant green-up in spring. The IRG is de-
calculated. Slope (degrees; 0–90), aspect (continuous degrees; 0–360,
fined as the first derivative of a double-logistic function fitted to the
where 0 is north and 180 is south), and elevation (m a.s.l.) were derived
annual time series of NDVI values scaled between 0 and 1 for a given
from a Digital Elevation Model (DEM). Aspect was further converted to
pixel, that is, when the change in NDVI value peaks. This metric has
northness (ranging from −1 to 1, where values close to −1 face south,
been verified by independent testing (Merkle et al., 2016). A space–
and values close to 1 face north) by cosine transformation. Distance to
time–time matrix that relates red deer movement data to the changes
outer coastline and distance to nearest fjord were measured in kilome-
in green-up in space and time was constructed for each individual
ters. In addition, the standard deviation of elevation was extracted for
deer. For each individual red deer, we calculated the cumulative IRG
each home range as a measure of the variation in topography.
(CIRG) over the entire growth season (1 April – 31 August) by summing
The proportions of different habitat types within the home ranges
the IRG for all pixels the animals used over the season at a given time.
were derived from digital land resource maps at a scale of 1:50,000
This represents the total instantaneous rate of green-up experienced
(Loe, Bonenfant, Meisingset, & Mysterud, 2012). The eight original
by an individual red deer throughout the growth season.
habitat types were simplified into four categories: pasture, forest, mountain (areas above the treeline), and all other habitat types (human settlement, marsh, water, glaciers, and areas not mapped). All land-
2.5 | Statistical analyses
scape maps were rasterized with a resolution of 100 m. In the model-
All statistical analyses were performed in R (R Development Core
ing, these variables were calculated as proportions within the seasonal
Team 2016). We used generalized linear mixed models in the library
home ranges of the individual deer.
lme4 (Bates & Maechler, 2009). The response variable was the CIRG
One may argue that summer home range characteristics are not
for the growth season. We used a random intercept for year to con-
the only relevant scale when measuring nutritional gain and that the
trol for annual variations in the mean CIRG due to climatic variation
annual range is also important. We therefore also calculated the dis-
and weather conditions. Sixteen observations (CIRG: n = 5; elevation:
tance between summer and winter ranges and the difference in el-
n = 1; Δelevation/distance summer-winter: n = 10) were removed due
evation between summer and winter ranges (Δelevation). The mean
to missing values in the covariates, leaving n = 330 individuals avail-
elevation of the winter and summer ranges was calculated based on
able for the analyses (182 migratory and 148 resident individuals).
the individual 95% fixed-kernel home ranges for March (n = 290), rep-
Marginal and conditional R2 were calculated following Nakagawa and
resenting winter, or April (n = 41) when data for March were not avail-
Schielzeth (2013).
able, and for July (n = 334), representing summer, or June (n = 5) when
For continuous variables, we log transformed or arcsine square
July data were not available. The choice of these months as summer
root transformed (habitat type, measured as proportions) covariates
and winter ranges corresponded well with red deer migration dates,
where appropriate to increase fit and stabilize the variance. Note
with a few exceptions for the winter range when individuals started
that although the use of arcsine square root transformation has been
spring migration toward the end of the chosen month (n = 8) or arrived
criticized (Warton & Hui, 2011), this is mainly in regard to its use for
in the summer range in the chosen month (n = 3). The centroids of in-
response variables and not for covariates as in our case. In addition,
dividual 95% minimum convex polygons from the same months were
some covariates were also rescaled by centering on the mean and di-
used to calculate the distance between summer and winter ranges, as
viding by the standard deviation where needed, that is, standardizing,
kernel home ranges often result in multiple polygons per individual,
as covariates being on very different scales causes model convergence
complicating centroid estimation.
issues. All covariates were assessed for nonlinearity with the response variable using GAM plots in the library mgcv (Wood, 2006), and ade-
2.4 | Plant phenology from satellite NDVI
quate parametrizations were chosen based on this. We also checked for correlations between all covariates, excluding the assumed least
The NDVI is a measure of the reflected photosynthetic activity in a de-
relevant one from the global model when r > |.6|. Categorical covari-
fined area (Pettorelli, Vik, et al., 2005) and is therefore commonly used
ates included in the model were sex and space use tactic (migratory
as a proxy for forage quantity and quality for ungulates (Garroutte,
or resident).
|
4451
MYSTERUD et al.
We used model selection with Akaike Information Criterion (AIC)
in benefits related to forage maturation (CIRG) included the landscape
to find the most parsimonious model (Burnham & Anderson, 2002).
variables elevation, proportion of forest and mountain, space use tac-
We considered models within ΔAIC